Proposer of the vote of thanks to Whiteley et al. and contribution to the Discussion of ‘Statistical exploration of the Manifold Hypothesis’

Whiteley et al. revisit one of the most influential ideas in modern statistical learning: the manifold hypothesis. While the hypothesis is routinely invoked to justify the success of high-dimensional data analysis, it is rarely given a probabilistic explanation. The central contribution of this paper is to show how low-dimensional geometric structure can emerge naturally from statistical dependence among random coordinate functions, rather than being imposed a priori. The proposed Latent Metric Model (LMM) provides a mathematically coherent framework linking latent structure, kernel geometry, and observed data geometry. In this sense, the paper offers not merely a reformulation of the manifold hypothesis, but a principled explanation for why manifold-like data structures may arise in practice.